AI Disclosure Statement
How AI involvement is recorded, what that supports under the EU AI Act, and where the limits genuinely are.
Position in one line. Cothenticity is not an AI detector. It is a disclosure instrument: it records what AI actually did, at the moment it did it, and publishes that alongside the work.
1. How AI involvement is recorded
When AI assistance inserts text in the Studio, that insertion is written to the event log as a
first-class ai_insert event carrying its position, length, and timestamp. Those events
are folded into the same SHA-256 hash chain as human keystrokes, so AI contribution cannot be
quietly removed from a receipt after the fact — deleting or shrinking one breaks verification, and
our server independently replays the chain and rejects any receipt that does not reconcile.
Every published receipt therefore states, and can substantiate, the proportion of the work that was human-typed, AI-assisted, and pasted.
2. EU AI Act, Article 50
Article 50 of Regulation (EU) 2024/1689 took effect on 2 August 2026, and requires that
certain AI-generated or AI-manipulated content be disclosed and marked in a machine-readable
format. Cothenticity is designed to support that obligation: each receipt exports a
machine-readable manifest containing an explicit cothenticity.ai.disclosure assertion
recording whether AI was used and to what extent.
Using Cothenticity does not by itself make you compliant. Article 50 obligations attach to providers and deployers of AI systems and depend on your own circumstances. Cothenticity provides evidence and a disclosure artefact; it does not constitute legal advice or a compliance certification. Take your own advice.
3. Content credentials
The exported manifest is aligned with the structure of C2PA / Content Credentials, which covers provenance at the media layer. Cothenticity covers the work layer — documents, drafts and code — which existing camera-and-pixel provenance does not reach. We intend to move toward full interoperability with that standard.
4. Honest limits
- We record the AI we can see. AI used inside the Studio is recorded precisely. Text produced in another tool and then pasted in is recorded as a paste, with its length and timing, but we cannot know its origin.
- Retyping defeats it. Text generated elsewhere and typed in by hand — or by an automated typer imitating human cadence — will be recorded as human. Defences against this are in development; they are not present in version 1.0.
- A receipt is evidence, not a verdict. It should inform a judgement, never replace one.
5. Our own use of AI
The Studio's built-in assistant is an AI feature, and its output is always attributed and disclosed in your receipt. We do not train models on your work. We do not read published receipts for any purpose other than operating and securing the Service.
6. Contact
[email protected]
Sam Squared Softwares (Pty) Ltd, registration number 2026/536077/07